view transcript

Lisa Su: The cloud data centers are very important to us. And when we look at it, you know, we had good adoption in the first generation of Epyc with Naples, you know, including Microsoft Azure and Amazon and, you know, Baidu and Tencent. I think what we see in the second generation of Epyc as it relates to both cloud and OEM is for those customers who were originally involved in the first generation of Naples, they're actually doubling down and they're increasing the number of instances. You know, you heard from HP enterprise yesterday that they're tripling, you know, their portfolio. And then, for some of the newer names, like for example, we're very proud of Google and their announcement. I think, you know, Google is, you know, just a marquee environment for, you know, let's call it tough, tough data center problems.

Lisa Su: And so to be able to be a part of that is a real honor for us. And, what you heard from Google is, you know, they have us in their production environments for internal workloads, as well as plans for Google, you know, cloud platform, later on in the year. You know, we're also very excited about the announcement with Twitter. You know, again, that's a, you know, let's call it an internal platform as it relates to, you know, cloud environments. And, I think we have both. So to your question of internal versus external, we have the external facing instances that you heard of from Microsoft Azure, and then we have, you know, some of the internal stuff that you've heard from, Google and Twitter

Eric Jhonsa: Along similar lines, given that the cloud giants, tend to be faster when it comes to adopting new silicon than like traditional enterprises. What portion of AMD's Rome sales do you expect, like in the next, say, two or three quarters to come from the cloud giants relative to the enterprise clients.

Lisa Su: I think the two largest segments for us over the next, you know, let's call it a couple of quarters to the, you know, the, the, the end of the year, will be cloud, you know, that's cloud, you know, sort of the, the tier one, you know, sort of the super seven as well as some of the, you know, next wave of cloud, companies as well as, HPC. We've done very, very well with HPC, a number of wins, with a number of OEMs and, these tend to ramp faster, because you know, they're a highly technical in their decision making. And once they, they make the decision, they move quickly, enterprise tends to take a little bit longer. You know, fortune 1000 CIOs take a bit longer in their evaluation cycle. We love the platform that we have, you know, if you look across HPE Dell, Lenovo, Supermicro, the ODM's, I think we have all the right platforms. It just takes a bit more time for them to complete their evaluations.

Eric Jhonsa: How do you see average selling for Rome trending relative to Naples going forward?

Lisa Su: Yeah, so we do see, so the way we think about this is look, we want to provide value to the data center operator. And you know, at the end of the day, there needs to be something in it for everyone, right? It is work, to switch from, you know, our competitor to AMD, there's work to be done, there's optimization to be done. And so, you know, we, we believe we're pricing at, let's call it excellent performance, at a fair price. And as it relates to, from our standpoint, you know, because the technology capability is going up, we do expect, you know, ASPs for us to increase, as we go from Naples to Rome as well. And, and that should reflect in our, you know, forward looking financials.

Eric Jhonsa: AMD has, over the last couple of years, talked up by its efforts to offer a lot of the features that Intel is traditionally reserved for dual socket servers on a single socket servers. And it seems with, with Rome, you're doubling down on that. Is there an estimate you can give for how much of AMD's second gen Epyc revenue could potentially involve single socket servers relative to dual socket servers? And are there particular types of customers that you feel are especially interested in them?

Lisa Su: Yeah, so, we do feel strongly about this. We do feel strongly about this idea that, customers should be able to decide, whether they want, you know, all processor performance or they want, you know, sort of the balance between processor and io. You know, I think single socket is still, new to many in the industry. And so, you know, we see more interest, but I would say it's still, it's still one of those areas where, you know, people have to get used to it. So I don't have an exact percentage today. I would say I expect it to grow as we go into, second generation Epyc compared to the first generation, but you know, the other thing about Rome is, you know, the, the dual socket performance is just incredible as well. And so, you know, we think both single socket and dual socket will be, will do well.

Eric Jhonsa: AMD has talked about, achieving a double digit servers CPU share within four to six quarters of achieving a mid single digit share, which you believe you achieved that by the end of 2018. Given how diverse the servers CPU market is, all the different types of customers and workloads there are, are there a particular market segments where you feel like AMD might be able to achieve, say, a 15% market share, you know, relative near future?

Lisa Su: Yeah. So as it relates to market share targets, I think we are, you know, we feel well positioned to, you know, achieve those, those targets that we previously talked about that would put, you know, sort of the double digit mark somewhere between sort of end of this year, middle of next year, and a lot of work has to happen, for that, but we feel good about that. As it relates to, you know, look what I say is look, 10% share is not a destination, right? It's a point in time. And, there are segments where we think our market share will be substantially higher than that. We certainly think in HPC the value proposition is very, very strong and we see that with, you know, both, you know, sort of all of, sort of the HPC providers. And we think in cloud, our value proposition is very, very strong. And I, I frankly, I think in enterprise is also quite strong. So it's really a matter of, sort of time to market and time to ramp. And, as I said earlier, I think, cloud and HPC move a little bit faster than traditional enterprise.

Eric Jhonsa: Yeah. Rome will be launched in about two years after AMD launched Naples at the same time, AMDs indicated it's a Nextgen Epyc platform codenamed named Milan will be available around mid 2020 and Intel for its part suggested it wants launched new servers CPU platforms once every four to five quarters. Does AMD have a target going forward regarding how frequently it wants to launch new servers CPU platforms?

Lisa Su: You know, I think we aim to be competitive. So if I, if I put that stake in the ground, you know, I think, you know, whether it's four or five or six quarters, it's in that range, but we aim to be very competitive.

Eric Jhonsa: I know it's early right now, but are there any details AMD can share about Milan's expected performance gains relative to Rome? Any like potential, what the potential improvements could be, instructions per clock or in what particular areas you, AMD especially wants to build on Romes advances?

Lisa Su: Well, I think it's a little early to talk about details of Milan. What I will say is that our zen three core design is complete. We like how it looks. We have plenty of ideas. Our engineers have plenty of ideas populating zen three and zen four, based on some of our learnings, off of Naples and Rome. And so, you know, we'll, we'll continue to push the envelope, but, you know, we'll talk more about it as we get closer to those products.

Eric Jhonsa: You know, I wanted to ask about graphics intensive workloads, in particular things like cloud gaming, virtual desktop, AMDs had some traction there. And obviously AMD has, you know, large GPU business. Is AMD interested in potentially developing servers CPUs with integrated GPUs?

Lisa Su: Great question. Look, I, first of all, I think the market around, you know, we love all things gaming. So when you think about gaming, you think about, you know, PC gaming, console gaming, you know, cloud gaming, they're all, they all need very powerful graphics, as well as processor capability. As it relates to, you know, cloud gaming in particular, we have had some success with, you know, Google and their Stadia launch. I think we're engaged across the industry. I wouldn't, I wouldn't, count anything out, you know, in terms of, we already have integrated graphics on our PC processors. You know, in server it's not so much whether it's integrated on the same chip that's important. What's really important is the connectivity between CPUs and GPUs. And so, we're very focused on our infinity architecture and how do we ensure that that connectivity is as efficient as it can possibly be.

Eric Jhonsa: You know, earlier this year AMD announced a deal Cray to supply CPUs and GPUs for the Department of Energy's frontier supercomputer. And, I thought it was interesting, you know, that the supercomputer will feature nodes that contain one custom AMD CPU that's connected to four GPUs through the infinity fabric. Is AMD open to offering something like that on an off the shelf basis rather than just on a custom basis.

Lisa Su: Yeah. So a great point. So we're extremely proud of, the, the frontier design win. I think it's one of those, you know, really big, tough problems that you have to put your best innovation together to make it happen. And it is a customized, you know, CPU and some custom GPU efforts for frontier to get to the, you know, over one and a half exaflops. I do think though, you're going to see many of those technologies, come into standard products. You know, as we go forward, you know, our goal is to push the envelope in HPC and so, you know, for, you know, general purpose, HPC applications, you'll see some of those technologies come into play

Eric Jhonsa: I wanted to ask about AMD's thoughts about Taiwan Semis Cemusa five nanometer manufacturing process. I know you mentioned that Milan will use seven nanometer plus process, but TSMC will be starting volume production for the five nm process in the first half of 2020. How is AMD broadly thinking about adopting five nanometer. For example, would it be a node that it would first deploy for like PCs, CPUs or GPUs as an in servers CPUs?

Lisa Su: Well, so we look at sort of technology intercepts very carefully. You know, we want to be early but not too early. You want to be, make sure that you can do a volume ramp. We haven't talked in detail about our five nanometer plans, just yet. I think, you know, TSMC is a great partner. We continue to work closely with them and you know, we'll talk more about, you know, our future technology plans later on.

Eric Jhonsa: I wanted to ask a little bit about AMD's plans for addressing the enterprise market. Like Nvidia's, you know, enter the hardware space, you know, with their DGX supercomputers, they have created their own systems. Given that AMD is, you know, has both server, CPUs and GPUs, would it be open to like launching its own hardware systems or would it prefer to just work with OEM partners?

Lisa Su: You know, we work really closely with OEM and ODM partners and that has been our go to market strategy. I think it's the right strategy. You know, we believe in this idea of an open ecosystem so you get the best of the best. And, you know, we continue to work on how do we specify those systems, so they take the best advantage of our hardware.

Eric Jhonsa: AMD's willingness to support next gen memory technologies, like Intel's talked up, its support for its Optane next gen memory, which relies on 3D cross point and that 3D crosspoint was co-developed with Micron, which plans to launch its own crosspoint product soon. Is AMD opened to partnering with a company like Micron to support next gen memory?

Lisa Su: Yeah. So, we are open to partnering across the ecosystem. So especially in memory, you know, it's so important to bring memory closer to the CPU and, and we know that that's one of the key things. You know, we have a great partnership with, you know, Micron, you know, Samsung as well as Hynix and , you know, we think pushing the envelope on that connection between CPU and memory is very important going forward.

Eric Jhonsa: Well, the AMD be open to supporting something like AVX 512 going forward?

Lisa Su: Well, so we did a tremendous amount of work around the floating point in zen 2. So we actually, you know, sort of doubled our floating point capability. And then once you look at the number of cores we've added, we've actually, you know, the, the, Rome architecture compared to Naples is about four times the floating point. We're always gonna look at how do we keep pushing on floating point and, and that, you know, again, that's, that's one of the mantras of, of leadership.

Eric Jhonsa: I wanted to ask about AMD's partnership with Xilinx. You know, you've been partnering with them for a while. The companies have a common competitor in Intel. Could you talk a little bit about, how you see that partnership potentially evolving and what types of workloads and customers you frequently see using both AMD, CPUs and Xilinx's FPGAs,

Lisa Su: Yeah. So, we are pleased with the partnership with Xilinx. I think the teams are very technically engaged and you know, most importantly we want to make sure that the interoperability, you know, we're both pushing the edge on the interconnect and so on, you know, PCI Gen four and, you know, all of these things. We're, we're very well aligned, in terms of, you know, early applications, you know, certainly there are some cloud environments that are looking at, you know, CPU plus FPGA environments. And so we have, some joint engagements in that area. You know, thinking about, you know, how, you know, some of these, you know, edge, you know, computing applications, you know, move forward. You know, we have some engagements in those areas as well.

Eric Jhonsa: You know, given that AMD's resources have been growing, is investing more in software developers and then like just a building out your engagements with third party developers and area of interest for the company?

Lisa Su: Totally. If you look at where we're adding resources, we're just adding a ton of resources in software and that software across, let's call it, you know, the basic, , yeah, sort of driver stack, which is really important for enablement, as well as, you know, software on, you know, libraries and also some of the machine learning frameworks, you know, ensuring that tensorflow runs well. Pytorch, you know, mxnet, you know, these guys. So I think, in our business, you know, the software is just so important to really enable the hardware

Eric Jhonsa: I wanted to ask also about Romes Chinese adoption. AMD mentioned during last week's earnings call that trade and tariffs uncertainties in the US entities list are affecting the sales a little bit. And how do you see Rome's adoption within the Chinese market Trending over the next few quarters?

Lisa Su: Yeah. So look, yeah, certainly we're all very cognizant of what's going on between the US and China and some of the tensions. You know, there are several of our customers that are on the entities list and, you know, we are not shipping to those customers. That being said, there's a broad market in China. I think Rome is extraordinarily competitive, whether again, you're talking about, you know, sort of cloud environments in the US or cloud environments in China as well as enterprise environments. And so we think Rome is gonna do well

Eric Jhonsa: Then I wanted to ask for a second about, like, interconnect technologies. AMD's, you know, supported a few different ones, you supported C6, supported [INAUDIBLE] and then more recently supported CXL, which was originally started by Intel. What's AMD's views on how these technologies will be adopted going forward and how they will coexist.

Lisa Su: Well, we feel really, you know, it's really important to ensure that the, the interconnect is open, so truly open. And that means, you know, no competitor has an advantage in, in any particular way. That being the case, you know, we are, you know, as you, as you mentioned, fairly open across standards and, you know, we continue to listen to, you know, sort of our customers and what they want and what they want is to be able to connect, you know, the best CPU to the best accelerator. And so that requires us to adopt, you know, a few technologies.

Eric Jhonsa: Your server CPU sales are grouped within the enterprise embedded and semi-custom segment, which also covers AMD's gaming console, processor sales. Given how the server's CPU business has been growing, is AMD open to like a separately breaking out its sales or potentially putting them in a new reporting segment that also covers server GPU sales?

Lisa Su: Yeah. So, yeah, we, we believe our segment reporting is, you know, the, the best way to do it right now for the company. We do try to provide additional color. I know there's a lot of interest around the data center. You know, I still consider ourselves very, very early, in the adoption cycle for a data center, in both CPUs and GPUs. And so, you know, we all will constantly look at this as we go forward, but we think these are the right reporting segments with the notion of, we will give, you know, additional color, to try to help the investment community.

Eric Jhonsa: All right. AMD hasn't historically bought many companies, but you know, its resources are growing, you know, you can potentially finance larger acquisitions, if you wanted it to either with cash or stock. How does AMD feel about potentially using M&A to strengthen its data center position and offer like a more, a platform level solutions there?

Lisa Su: Yeah. So, I think that it is definitely something that we look at and think about. You know, we've done some, smaller scale M&A to acquire, you know, mostly skills, in recent years and those aren't things that, you know, we talk, you know, in a big way about, and I think as we go forward, you know, as the balance sheet strengthens, as the revenue grows, as, you know, we have, you know, more capability, you know, we'll certainly look at M&Aas a, as a way to accelerate the business.

Eric Jhonsa: All right. And my last question was about your future plans. Earlier this week you strongly denied a rumor about leaving for IBM and you've been AMD CEO for almost five years now. I guess my question would be how long would you see yourself potentially staying at AMD and are there particular longterm goals that you're looking to see through as AMD CEO?

Lisa Su: Well, I can say for sure I'm staying at AMD. So, look, I, you know, one of the things that, that is really important to me is, look, we are still in the early innings of the AMD story. You know, if you take a look at how long it takes for a tech company to really complete a full cycle, you know, although I've been CEO for five years, we're just at the cusp of some incredibly competitive products. And, you know, I'd like to see us become one of the premier growth franchises in tech. And you know, that means growing market share and servers and PCs and continuing to deepen our partnerships. So, yeah, I have a lot to do.

Lisa Su says she has no plans to leave AMD (AMD - Get Report)  and that there's much that she still wants to accomplish as the CPU and GPU developer's CEO.


During a wide-ranging conversation with TheStreet that took place a day after AMD held a very well-received launch event for its second-generation Epyc server CPUs, Su forecast the CPUs, code named Rome, will see strong early adoption within cloud servers and high-performance computing (HPC) systems. She also discussed AMD's landing of Google (GOOGL - Get Report) as a Rome client, partnerships with Xilinx (XLNX - Get Report) and memory makers, expanding investments in software and her thoughts on using acquisitions to strengthen AMD's competitive position.

And following her denial of a recent report stating that she's thinking about leaving AMD for IBM, Su indicated she'll be at AMD for a while to come.

"I can say for sure, I'm staying at AMD," Su said when asked about her future plans. "Although I've been CEO for five years, we're just at the cusp of some incredibly competitive products. And I'd like to see us become one of the premier growth franchises in tech...So yeah, I have a lot to do."  

 

Notable Quotes from the CEO

On Cloud Giants:

what you heard from Google is, you know, they have us in their production environments for internal workloads, as well as plans for Google, you know, cloud platform, later on in the year. You know, we're also very excited about the announcement with Twitter. You know, again, that's a, you know, let's call it an internal platform as it relates to, you know, cloud environments. And, I think we have both.

Achieving Duble Digit Share:

I think we are, you know, we feel well positioned to, you know, achieve those, those targets that we previously talked about that would put, you know, sort of the double digit mark somewhere between sort of end of this year, middle of next year, and a lot of work has to happen, for that, but we feel good about that.

Trade Tensions with China:

certainly we're all very cognizant of what's going on between the US and China and some of the tensions. You know, there are several of our customers that are on the entities list and, you know, we are not shipping to those customers.

Cloud Gaming:

in server it's not so much whether it's integrated on the same chip that's important. What's really important is the connectivity between CPUs and GPUs. And so, we're very focused on our infinity architecture and how do we ensure that that connectivity is as efficient as it can possibly be.

 AMD's Partnership with Xilinx:

We're, we're very well aligned, in terms of, you know, early applications, you know, certainly there are some cloud environments that are looking at, you know, CPU plus FPGA environments. And so we have, some joint engagements in that area.

Future Plans with AMD:

Well, I can say for sure I'm staying at AMD. So, look, I, you know, one of the things that, that is really important to me is, look, we are still in the early innings of the AMD story... So, yeah, I have a lot to do.

Premium Pick: Jim Cramer: Fear Not These False Market Fears

Ask the Expert: Cannabis One CEO Believes Now Is the Time to Invest in Cannabis

Ready to Retire: The Biggest Threat to Your Retirement? Check Your Basement

TheStreet Feature: Here's Something Investors May Be Missing About the Drone Revolution

Dog Days of Summer: Why Investors May Be Seeing the End of the Dog Days of Summer

Trivia Time: Guess This Landmark Named After One of Our Founding Fathers

Catch Up: Today's Top News Videos Below